Estimation of all-sky daily air temperature with high accuracy from multi-sourced data in China from 2003 to 2022

A long, high-quality, and temporally continuous high spatiotemporal resolution air temperature (Ta) dataset plays a crucial role across various domains, particularly in areas such as human health, disease prediction and control, and energy utilization, where extreme temperatures (daily maximum and minimum temperatures) hold significant value. However, due to the instability of extreme temperatures influenced by various factors like topography, altitude, climate, and underlying surfaces, coupled with sparse meteorological station coverage, traditional methods struggle to accurately capture and produce high-quality, temporally continuous temperature dataset products. In this study, the four-dimensional spatiotemporal deep forest (4D-STDF) model was utilized, based on daily meteorological station temperature data from 2003 to 2022, along with seamless daily LST, meteorological, radiational, land use, topographic and population data encompassing 12 parameter factors and 6 spatiotemporal factors, three high-quality daily Ta datasets were constructed and generated. These datasets cover mainland China, featuring high spatial resolution (1km), long temporal sequences (2003-2022), and increased accuracy. The datasets include maximum (T_max), minimum (T_min), and mean (T_mean) temperatures from January 1, 2003, to December 31, 2022, as well as monthly and yearly synthesized T_max, T_min, and T_mean values, presented in GeoTIFF format with WGS84 projection, and the data unit is in 0.1 degrees Celsius (°C). The overall RMSE values are 1.49°C, 1.53°C, and 1.18°C for daily estimates, 1.38°C, 1.65°C, and 0.52°C for monthly, and 1.28°C, 1.83°C, and 0.41°C for annual, respectively. These datasets reasonably capture the spatial and temporal heterogeneity of Ta and effectively capture the intensity of heatwaves and cold spells. These new datasets are of significant value for studying extreme climates and contribute to assessing their impact on human health, infrastructure, and energy demands.